A shift for cognitive neuroscience: from model-based data to data-driven models

The brain has a complex organization portraying multiple topographies at different levels and along different axes. Following the developments of neuroimaging technics in the last decades, different approaches have been developed to study brain organization based on neuroimaging features. Furthermore, many concepts from the psychological sciences have been mapped to the brain. Thousands of fMRI and PET studies show the neural correlates of various concepts derived from the study of human behavior. An overview of those model-based data reveals a many-to-many mapping on both brain-behavior directions: one single brain region or network can be related to many behavioral concepts and conversely, one single behavioral concept can be assigned to many brain regions/networks. The aggregation of these behavioral model-based data can now be harnessed to switch the perspective. By using quantitative approaches on available neuroimaging data, we can derive “behavioural profiling” of brain regions and networks. Such data-driven approaches can be used, in turn, to generate a brain-based model of cognitive information processing.